American Journal

Journal of Science and Technology

2014; 1(1): 36-49
Published online March 30, 2014 (http://www.aascit.org/journal/ajst)

The process of heat transfer and

fluid flow in CFD problems
Anestis Stylianos
Department of Oenology and Beverage Technology, Technological Educational Institute
of Athens, Faculty of Food and Nutrition, Greece

Email address
sanestis@teiath.gr, Anestis@essolutions.eu

Citation
Anestis Stylianos. The Process of Heat Transfer and Fluid Flow in CFD Problems.
American Journal of Science and Technology. Vol. 1, No. 1, 2014, pp. 36-49

Keywords Abstract
CFD, In many applications we need to use the CFD technology in order to model the
Heat Transfer, problem and find exact solutions with less cost. In market can be found many
Fluid Flow, special CFD software solutions where the user must programming according its
SIMPLE Algorithm problem and its special situations he wants to solve. In order to formulate the
starting problem in a suitable form acceptable from the CFD package must
follow a variety of steps which are no so clearly. Here, we try to present all these
steps, and give the basic equations per step.
Received: March 03, 2014
Revised: March 17, 2014 1. Introduction
Accepted: March 18, 2014
Importance of heat transfer and fluid flow. This paper is concerned with heat
and mass transfer, fluid flow, chemical reaction, and other related processes that
occur in engineering equipment, in the natural environment, and in living
organisms. That these processes play a vital role can be observed in a great
variety of practical situations. Nearly all methods of power production involve
fluid flow and heat transfer as essential processes. The same processes govern
the heating and air conditioning of buildings. Major segments of the chemical
and metallurgical industries use components such as furnaces, heat exchangers,
condensers, and reactors, where thermo fluid processes are at work. Aircraft and
rockets owe their functioning to fluid flow, heat transfer, and chemical reaction,
(Sunden, 2013). In the design of electrical machinery and electronic circuits,
heat transfer is often the limiting factor. The pollution of the natural environment
is largely caused by heat and mass transfer, and so are storms, floods, and fires.
In the face of changing weather conditions, the human body resorts to heat and
mass transfer for its temperature control. The processes of heat transfer and fluid
flow seem to pervade all aspects of our life.
The purpose of this paper is primarily to aim at developing a general method
of prediction for heat and mass transfer, fluid flow, and related processes. As we
shall shortly see, among the different methods of prediction, the numerical
solution offers great promise. We shall present a numerical method having
complete generality for predicting the processes of interest. We shall, therefore,
refrain from accepting any final restrictions such as two-dimensionality,
boundary-layer approximations, and constant-density flow. If any restrictions are
temporarily adopted, it will be for ease of presentation and understanding and
not because of any intrinsic limitation. Numerical solution therefore becomes a
desirable approach for heat conduction under non-linear, complex geometric
configurations, or complex boundary conditions. Instead of obtaining the
 American Journal of Science and Technology 2014, 1(1): 36-49
analytical expression of the temperature distribution, the
results of numerical solutions are given at discrete points.
The numerical solution involves three steps: (1)
discretization of the computational domain, (2)
discretization of the governing equations, and (3) solution
of the algebraic equations (Murthy et al., 2006).
1.1. Basic Definitions
Prediction of heat transfer and fluid-flow processes can
be obtained by two main methods: experimental
investigation and theoretical calculation. We shall briefly
consider each and then compare the two.
Experimental Investigation. The most reliable
information about a physical process is often given by
actual measurement. An experimental investigation
involving full-scale equipment can be used to predict how
identical copies of the equipment would perform under the
same conditions. Such full-scale tests are, in most cases,
prohibitively expensive and often impossible. The
alternative then is to perform experiments on small-scale
models. The resulting information, however, must be
extrapolated to full scale, and general rules for doing this
are often unavailable. Further, the small-scale models do
not always simulate all the features of the full-scale
equipment; frequently, important features such as
combustion or boiling are omitted from the model tests.
This further reduces the usefulness of the test results.
Finally, it must be remembered that there are serious
difficulties of measurement in many situations, and that the
measuring instruments are not free from errors,
[Hirsch,2007].
Theoretical Calculation. A theoretical prediction works
out the consequences of a mathematical model, rather than
those of an actual physical model. For the physical
processes of interest here, the mathematical model mainly
consists of a set of differential equations. If the methods of
classical mathematics were to be used for solving these
equations, there would be little hope of predicting many
phenomena of practical interest. A look at a classical text
on heat conduction or\ fluid mechanics leads to the
conclusion that only a tiny fraction of the range of practical
problems can be solved in closed form. Further, these
solutions often contain infinite series, special functions,
transcendental equations for eigenvalues, etc., so that their
numerical evaluation may present a formidable task,
(Patankar, 1975).
We shall now list the advantages that a theoretical
calculation offers over a corresponding experimental
investigation. These are low cost, speed, complete
information, ability to simulate realistic and ideal
conditions. The foregoing advantages are sufficiently
impressive to stimulate enthusiasm about computer
analysis. A blind enthusiasm for any cause is however
undesirable. It is useful to be aware of the drawbacks and
limitations. For the purpose of discussing the disadvantages
of a theoretical calculation, it is, therefore, useful to divide
37
all practical problems into two groups:
Group A: Problems for which an adequate mathematical
description can be written. (Examples: heat conduction,
laminar flows, and simple turbulent boundary layers.). For
difficult problems involving complex geometry, strong
nonlinearities, sensitive fluid-property variations, etc., a
numerical solution may be hard to obtain and would be
excessively expensive if at all possible. Extremely fast and
small-scale phenomena such as turbulence, if they are to be
computed in all their time-dependent detail by solving the
unsteady Navier-Stokes equations, are still beyond the
practical reach of computational methods.
Group B: Problems for which an adequate mathematical
description has not yet been worked out. (Examples:
complex turbulent flows, certain non-Newtonian flows,
formation of nitric oxides in turbulent combustion, some
two-phase flows.) This research consists of proposing a
model, working out its implications by computer analysis,
and comparing the results with experimental data. Thus,
computational methods play a key role in this research.
The steps which must follow someone in order to use
CFD software for the suitable solution to his problem can
be presented next, Fig 1. These are:
Fig 1. The use of CFD Software.
1. Specify the basic problems equations in 1D, 2D or
3D space. These are the mass, the momentum, the
energy, the molecular and the temperature ones.
2. Specify the turbulence model if you have.
3. Solve the above equations and modify them so as
their final form will take the general form of
equation 8. Your scope will be to specify all the
appropriate coefficients to input in the software.
4. Discretize the RANS equations following the finite
elements or the Boundary elements or the Finite
difference method.
1. Specify the grid according to its geometry
2. Choose the appropriate scheme
3. Specify the over relaxation and/or under
relaxation factors
4. Specify the convection and/or diffusion
38 Anestis Stylianos: The Process of Heat Transfer and Fluid Flow in CFD Problems

factors Let ml denote the mass fraction of a chemical species. In

5. Specify the pressure and the velocity
corrections
1. Specify initial values and boundary conditions
2. Choose a computer algorithm
3. Choose a CFD software program.
2. Basic Equations
The numerical solution of heat transfer, fluid flow, and
other related processes can begin when the laws governing
these processes have been expressed in mathematical form,
generally in terms of differential equations. For a detailed
and complete derivation of these equations, the reader
should turn to a standard textbook. Our purpose here is to
develop familiarity with the form and the meaning of these
equations. It will be shown that all the equations of
relevance here possess a common form, the identification
of which is the step toward constructing a general solution
procedure, (Lauder and Spalding, 1972).
the presence of a velocity field u, the conservation of ml is
expressed as:
(2)
Where l is the diffusion coefficient,
denotes the rate of change of the mass of the chemical
species per unit volume. Also, the quantity uml is the
convection flux of the species, the Rl is the rate of
generation of the chemical species per unit volume
(negative, zero or positive,(Launder and Spalding, 1974).
The energy equation in its most general form contains a
large number of influences. Since we are primarily
interested in the form rather than in the details of the
equation, it will be sufficient to consider some restricted
cases. For a steady low-velocity flow with negligible
viscous dissipation, for ideal gases, with c as a constant
pressure specific heat and h = cT, the energy equation can
be written as
" &'
! % (3)
# #

The differential equation governing the conservation of

momentum in a given direction for a Newtonian fluid can
be written along similar lines; however, the complication is
greater because both shear and normal stresses must be
considered and because the Stokes viscosity law is more
complicated than Fick's law or Fourier's law. With u
denoting the x-direction velocity, we write the
corresponding momentum equation as
*
Fig 2. Flux balance over a control volume. ( ) + , (4)

The individual differential equations that we shall The momentum equation can be written in Cartesian
encounter express a certain conservation principle. Each coordinate system as:
equation employs a certain physical quantity as its
* 0
dependent variable and implies that there must be a balance - ) !(.// %
0 0 0 2
!(.// % !(.// % !( 1 % !( 1 %
among the various factors that influence the variable. The
dependent variables of these differential equations are
3
usually specific properties, i.e., quantities expressed on a !( 1 % (5)
unit-mass basis. Examples are mass fraction, velocity (i.e.,
* 2
momentum per unit mass), and specific enthalpy. The - ) !(.// %
terms in a differential equation of this type denote 2 2 0 2
influences on a unit-volume basis. An example will make !(.// % !(.// % !( 1 % !( 1 %
3
!( 1 %
this clear. Suppose J denotes a flux influencing a typical
(6)
dependent variable . Let us consider the control volume of
dimensions dx, dy, dz shown in Fig. 2. So, we can take the
- - --
4 5 6
net balance per unit volume:
7 - - -
(1) ) 8(.// 9 8(.// 9 8(.// 9
6 4 4 5 5 6 6
0 2 3
Another example of a term expressed on a unit-volume !( 1 % !( 1 % !( 1 % (7)
basis is the rate-of-change term . If is a
specific property and is the density, then denotes the The density may be related, via an equation of state, to
amount of the corresponding extensive property contained variables such as mass fraction and temperature. These
in a unit volume. Thus, is the rate of change of variables and the velocity components obey the general
the relevant property per unit volume. differential equation. Further, the flow field should satisfy
 American Journal of Science and Technology 2014, 1(1): 36-49
an additional constraint, namely, the mass conservation or
the continuity equation, which is
: The numerical procedures which are associated with
t div u 0 (8)
:;
Another useful representation is the Cartesian-tensor
form of these equations:
C
E FG 0
D Integral methods involve the integral parameters of the
Where the subscript j can take the values 1, 2, 3,
denoting the three spaces coordinates. When a subscript is
repeated in a term, a summation of three terms is implied;
for example,
E FG
J L
D H I K
An immediate benefit of the Cartesian-tensor form is
that the one-dimensional form of the equation is obtained
by simply dropping the subscript j. The procedure for
casting any particular differential equation into the general
form (16) is to manipulate it until, for the chosen
dependent variable, the unsteady term and the convection
and diffusion terms conform to the standard form. The
coefficient of grad in the diffusion term is then taken as
the expression for , and the remaining terms on the right-
hand side are collectively defined as the source term S.
3. The Turbulence Model
The equations for unsteady laminar flow are converted
into the time averaged equations for turbulent flow by an
averaging operation in which it is assumed that there are
rapid and random fluctuations about the mean value. The
additional terms arising from this operation are the so-
called Reynolds stresses, turbulent heat flux, turbulent
diffusion flux, etc. To express these fluxes in terms of the
mean properties of the flow is the task of a turbulence
model.
Many turbulence models employ the concept of a
turbulent viscosity or a turbulent diffusivity to express the
turbulent stresses and fluxes. The result is that the time-
averaged equations for turbulent flow have the same
appearance as the equations for laminar flow, but the
laminar exchange coefficients such as viscosity, diffusivity,
and conductivity are replaced by effective (i.e., laminar
plus turbulent) exchange coefficients. From a
computational viewpoint, a turbulent flow within this
framework is equivalent to a laminar flow with a rather
complicated prescription of viscosity. (The same idea is
applicable to non-Newtonian flows, which can be thought
of as flows in which the viscosity depends on the velocity
gradient. The currently popular "two-equation models" of
turbulence (Launder and Spalding, 1972, 1974) employ, as
one of the equations, the equation for the kinetic energy k
of the fluctuating motion and the term (G-) is the net
source term, then we take:
39
M M " M N ) O (11)
turbulence models to make complete calculation methods
can be divided into integral and differential types.
Differential methods involve direct assumptions for the
Reynolds stresses at a point and seek the solution of the
governing equations in their partial differential form.
(9)
shear layer momentum thickness, shape parameter, skin
friction coefficient, etc. One solves a system of ordinary
differential equations (for 2-D flows), whose dependent
variables are the profile parameters and independent
variable is x; in 3-D flows, the equations are the partial
differential equations in the plane of the layer. The
(10)
important distinction between calculation methods is the
type of turbulence model rather than the type of numerical
procedure. The advantage of differential methods is that the
restrictions and inaccuracy that arise from the need to
parameterize the velocity profiles are avoided. Differential
methods introduce substantially more detailed information
about turbulence.
The turbulence models can be classified in several ways.
The one most often used is that arranged in order of the
number of differential equations solved in addition to the
mean flow equations [Reynolds, 1976].
(I) Zero equation models
(II) One equation models
(III) Two equation models
(IV) Stress equation models
Most of the models, classes (I)-(III), use Boussinesq
eddy viscosity model. Bradshaw et al. (1967), however,
assume the constancy of the r/pk ratio, in boundary layer
flows. Here z is the shear stress and k is the turbulent
kinetic energy. It is important to note here that in this case
the mean momentum and continuity equations form a
hyperbolic system in contrast to the parabolic system
obtained with the use of eddy viscosity models [Cebeci et
al., 1980]. Other models which do not use the eddy
viscosity assumption (class IV) obtain the Reynolds stress
from a differential equation.
Zero equation model, which uses only the partial
differential equation for the mean flow field and no
transport equations for turbulence quantities, is also called
"mean field" closure [Mellor and Herring, 1973]. The
classes (II) to (IV) are called "transport equation" closures.
Bradshaw (1972) describes the interplay between the
development of models and the experiments.
Turbulence models for general-purpose CFD must be
frame-invariant i.e. independent of any particular
coordinate system and hence must be expressed in tensor
form. These rules out simpler models of boundary-layer
type (e.g. mixing-length models). Turbulent flows are
computed either by solving the Reynolds-averaged Navier-
Stokes equations with suitable models for turbulent fluxes
or by computing the fluctuating quantities directly. The
main approaches are summarized below.
40 Anestis Stylianos: The Process of Heat Transfer and Fluid Flow in CFD Problems
Reynolds-Averaged Navier-Stokes (RANS) Models
Eddy-viscosity models (EVM)
assume the (deviatory) turbulent stress
proportional to the mean rate of strain;
Eddy viscosity derived from turbulent
transport equations (usually k + one other).
Non-linear eddy-viscosity models (NLEVM)
turbulent stress modelled as a non-linear
function of mean velocity gradients;
turbulent scales determined by solving
transport equations (usually k + one other)
Mimic response of turbulence to certain
important types of strain.
Differential stress models (DSM)
aka Reynolds-stress transport models (RSTM)
or second-order closure (SOC);
Solve transport equations for all turbulent
stresses.
Computation of fluctuating quantities
Large-eddy simulation (LES)
Compute time-varying flow, but model sub-
grid-scale motions.
Direct numerical simulation (DNS)
No modelling; resolve the smallest scales of
the flow.
The monograph by Launder and Spalding (1972) gives
the mathematical concepts of turbulence models. Bradshaw
and Cebeci (1978) describe the calculation methods for
various classes of turbulent flows. Gosman et al. (1969)
present various aspects of computation of recirculating
flows. Several reviews have appeared concentrating on
different aspects of turbulence modeling (Lauder and
Spalding, 1972; Rotta, 1975; Lumley, 1978; Rodi, 1982;
Spalding, 1982). A general review of turbulence models
and their applications can be found in Refs [Lumley, 1978;
Rodi, 1982; Spalding, 1982; Launder and Spalding, 1974;
Meller and Yamada, 1982]. Launder and Spalding (1972)
review briefly the turbulence models and their applications
to internal flows, concentrating on their work till 1972. The
reviews by Marvin (1983), Murphy (1984) and Wilcox and
Rubesin (1980) have concentrated on turbulence models
and their applications to external flows. In particular, they
concentrated on the models used by NASA Ames groups.
Lumley (1978, 1983) and Launder (1982) discuss the
prospects of higher order closure models. Turbulence
models and their application to atmospheric turbulence are
reviewed by Donaldson (1972). Launder et al (1984)
discuss second-moment closures in great detail. Meller and
Yamada (1982) described the development and application
of closure models to atmospheric and oceanic boundary
layers. Rodi (1984) describes in detail the various
turbulence models and their application to hydraulics
problems. He evaluates the models with regard to their
predictive capability and computational effort. Turbulence
models available for the prediction of complex 3-D flows
with curvature, rotation and flow separation are reviewed
by Lekshminarayana (1985). Patel et al.(1985) provide an
exhaustive review of turbulence models for near-wall and
low Reynolds number flows. The present review
concentrates on turbulence models and their applications to
a number of internal flows. The flows considered include
2- and 3-D flows, flows with and without swirl and flows
with secondary motion. In the following, the main classes
of turbulence models are described briefly.
The standard high Reynolds number k- turbulence
model has been widely used in industrial applications to
predict the overall performance of a device. The model has
been proved to be very robust and economical from the
view point of computer time because of the use of standard
wall functions. However, it has been observed that in
recirculating flow, the prediction of near wall quantity
using the k- model does not compare very well with other
low Reynolds number models. So, for the accurate
prediction of overall quantity (the mean residence time,
mixed volume and dead volume in a situation where
mixing is of importance) in a device, modified forms of the
standard k- model have been developed in the last decade.
However, such modified k- models has not been used very
extensively for industrial cases except its validation with
simple experiments. It has been the main motivation of the
present work to use the standard k- model of Launder and
Spalding (1972) along with its modifications, RNG
(Yahkot and Orszag, 1992), Chen-Kim (CK) and k- with
Yap correction (k- Yap) (Monson et al., 1990). The four
turbulence models listed above are all of high Reynolds
number form and are, of course, restricted to situations in
which the Reynolds number is sufficiently high for the
viscous effect to be unimportant. However, to describe the
flow close to a solid wall where the Reynolds number is
not sufficiently high, low Reynolds number turbulence
models with near-wall modifications have been reported to
perform better (Patel et al., 1986). In many computations,
we intend to use various low Reynolds number models like
the Lam-Bremhorst model (LB model) (Lam and
Bremharet, 1981), the Chen-Kim low Reynolds number
(CK low Re) model (with and without Yap correction) and
the simplest of the turbulence models the constant effective
viscosity model.
Conservation equation for the turbulence kinetic energy
PQRR "
M M -M ! %
ST
PQRR " PQRR "
! % ! % N" ) O (12)
ST ST
Conservation equations for the dissipation rate of
turbulence kinetic energy
PQRR "
M M -M ! %
ST
PQRR " PQRR "
! % ! % N" ) O (13)
ST ST
In these equations Gk represents the generation of
turbulence kinetic energy due to the mean velocity
gradients and can be expressed as;
 American Journal of Science and Technology 2014, 1(1): 36-49 41

2 2 - 2 2 - 2
N" 2( V W! % ! % ! % X W! 5 % ! % model equations.
4 5 6 4 6 4
- 2
! % Y
Table 1. Coefficients value for the equations 16.
6 5
(14)
Equation i S
The turbulent or eddy viscosity T, is computed by Mass 1 0 0
Chemical species mi -i gradmi Ri
7
combining k and as follows:
) + ,
4
Momentum u,v,w gradu
#Z C" I
(1 (15)
[
Energy Ti K/c gradT Sh/c
f
Turbulence (
g"
k (G-)
Finally, c1, c2, c, k, are empirical constants with kinetic energy
Dissipation rate f M
(
standard values [Launder and Spalding, 1974] as:
g[ h N ) h.J O
O .
of turbulence
C1 = 1.43, c2 = 1.92, c = 009, k = 1.0, = 1.3 kinetic rate

Another useful representation is the Cartesian-tensor

4. The General Form of Equations form of these equations. Where the subscript j can take the
values 1, 2, 3, denoting the three spaces coordinates. When
This brief journey through some of the relevant
a subscript is repeated in a term, a summation of three
differential equations has indicated that all the dependent
terms is implied; for example,
variables of interest here seem to obey a generalized
conservation principle. If the dependent variable is denoted
by =(x,y,z,t) the general differential equation is (20)

div u div grad S (16)

:;
b E F bG
D D
The four terms in the general differential equation (16)
are the unsteady term, the convection term, the diffusion
term, and the source term. The dependent variable can
stand for a variety of different quantities, such as the mass
fraction of a chemical species, the enthalpy or the
temperature, a velocity component, the turbulence kinetic
energy, or a turbulence length scale. Accordingly, for each
of these variables, an appropriate meaning will have to be
given to the diffusion coefficient rand the source terms. -
Not all diffusion fluxes are governed by the gradient of the
relevant variable. The use of div(grad) as the diffusion
term does not, however, limit the general equation to
gradient-driven diffusion processes. Whatever cannot be
fitted into the nominal diffusion term can always be
expressed as a part of the source term; in fact, the diffusion
coefficient can even be set equal to zero if desired. A
gradient-diffusion term has been explicitly included in the
general equation because most dependent variables do
require a prominent diffusion term of this nature.
Many times we can rewrite the eq.16 in equal definitions
forms like:
div u ) grad
:;
div u
:;
In eq.19 we replace the diffusion term by the Fick law
equation where J=-grad. Many times in Fick law we use
the diffusion coefficient D = /. Using the eq. 16 and
changing the values in , we can take the mass, the
momentum, the energy, the temperature, the turbulence
Definitions. A two-way coordinate is such that the
conditions at a given location in that coordinate are
c
8 9 d (17) influenced by changes in conditions on either side of that
D
location. A one-way coordinate is such that the conditions
at a given location in that coordinate are influenced by
changes in conditions on only one side of that location,
(Patankar and Spalding, 1970).
Parabolic, elliptic, hyperbolic. It appears that the
mathematical terms parabolic and elliptic, which are used
for the classification of differential equations, correspond
to our computational concepts of one-way and two-way
coordinates. The term parabolic indicates a one-way
behavior, while elliptic signifies the two-way concept. A
situation is parabolic if there exists at least one one-way
coordinate; otherwise, it is elliptic. A flow with one one-
way space coordinate is sometimes called a boundary-
layer-type flow, while a flow with all two-way coordinates
is referred to as a recirculating flow (Patankar and Spalding
1970; Gosman at al., 1969).
Discretization. For a given differential equation, the
required discretization equations may be derived in many
ways such as Taylor Series Formulation, Variational
Formulation, Method of weighted Residuals, Control
Volume Formulation, etc. The result of the discretization
must satisfy the four basic rules, which are a) the
consistency at control volume faces, b) the positive
S (18) coefficients, c) the negative slope linearization of the
source terms and d) the sum of the neighbor coefficients.
J S (19) Steady one-dimensional
We have already derived the discretization equation for
steady conduction in one dimension. The governing
differential equation, where with S we show the rate of heat
generation per unit volume, Fig. 3, is:
i i1
!M % d 0 (21)
i i
42 Anestis Stylianos: The Process of Heat Transfer and Fluid Flow in CFD Problems
This leads to the discretization equation
(22)
Unsteady one-dimensional
With reference to the general differential equation for ,
we have now seen, at least in the one-dimensional context,
how to handle the diffusion term and the source term. Here,
we turn to the unsteady term and temporarily drop the
source term, since nothing new needs to be said about it.
Thus, we seek to solve the unsteady one-dimensional heat-
conduction equation
(23)
Fig 3. Grid point cluster for the one dimensional problem.
The discretization equation is now derived by integrating
Eq. (23) over the control volume shown in Fig.3 and over
the time interval from t to (t+t). Thus
(24)
After representation of some terms, assumptions and
replacements we take as final result:
(25)
Discretization Equation for Two Dimensions
A portion of a two-dimensional grid is shown in Fig.4.
For the grid point P, points E and W are its x-direction
neighbors, while N and S (denoting north and south) are
the y-direction neighbors. The control volume around P is
shown by dashed lines. Its thickness in the z direction is
assumed to be unity. The nomenclature introduced in Fig.3
for distances x, (x)e, etc. is to be extended to two
dimensions here. Locating them exactly midway between
the neighboring grid points is an obvious possibility, but
other practices can also be employed. We have seen how to
calculate the heat flux Qe at the control-volume face
between P and E. We shall assume that Qe, thus obtained,
prevails over the entire face of area (y*1). Heat flow rates
through the other faces can be obtained in a similar fashion.
In this manner, the differential equation
(26)
Can be instantly turned into the discretization equation
(27)
Discretization Equation for Three Dimensions
Finally, we add two more neighbors T and B (top and
bottom) for the z direction to complete the three-
dimensional configuration. The discretization equation can
easily be seen to be
(28)
Fig 4. Control volume for the two dimensional problem.
4.1. The Use of a Grid
Finite difference equations are most efficiently solved in
a rectangular domain (for 2-D applications and an
equivalent hexahedral domain for 3-D applications) with
equal grid spacing. Unfortunately, the majority of physical
domains encountered are nonrectangular in shape. Thus, it
is necessary to transform the nonrectangular physical
domain to a rectangular computational domain where grid
points are distributed at equal spacing. It is also important
to note that the transformation allows the alignment of one
of the coordinates along the body, thus facilitating the
implementation of the boundary conditions. The objective
of grid generation is then to identify the location of the grid
points in the computational domain and the location of the
corresponding grid points in the physical space.
Furthermore, the metrics and Jacobian of transformation
which are required for the solution of flow equations are
computed within the grid generation routine, [Hoffman and
Chiang, 2000].
Typically, grid generation schemes may be categorized
as algebraic methods or differential methods. In the latter
case, the scheme is based on the solution of a set of PDEs
and may be subcategorized as either an elliptic, parabolic,
or hyperbolic grid generation. Either category of grid
generation scheme should include the following
 American Journal of Science and Technology 2014, 1(1): 36-49 43

considerations.
1. A mapping which guarantees one-to-one (30)
correspondence ensuring grid lines of the same family do
not cross each other;
2. Smoothness of the grid distribution; (31)
3. Orthogonality or near orthogonality of the grid lines;
4. Options for grid clustering.
A brief summary of the advantages and disadvantages of (32)
each method is provided below.
1. Algebraic grids
The advantages of this category of grid generators are: (a) (33)
They are very fast computationally; (b) Metrics may be
evaluated analytically, thus avoiding numerical errors; (c) (34)
The ability to cluster grid points in different regions can be
easily implemented.
The disadvantages are: (a) Discontinuities at a boundary
may propagate into the interior region which could lead to
errors due to sudden changes in the metrics; (b)
Smoothness and skewness may be difficult to control.
2. Elliptic grids
The advantages of this class of grid generators are: (a)
Will provide smooth grid point distribution, i.e., if a
boundary discontinuity point exists, it will be smoothed out
in the interior domain; (b) Numerous options for grid
clustering and surface orthogonality are available; (c)
Method can be extended to 3-D problems.
The disadvantages of the method are: (a) Computation
time is large (compared to algebraic methods or hyperbolic
Fig 5. Variation of temperature with time for different schemes.
grid generators); (b) Specification of the forcing functions
P and Q (or the constants used in these functions) is not 4.3. Over Relaxation and under Relaxation
easy; (c) Metrics must be computed numerically.
3. Hyperbolic grids In the iterative solution of the algebraic equations or in
The advantages of hyperbolic grid generators are: (a) the overall iterative scheme employed for handling
The grid system is orthogonal in two dimensions; (b) Since nonlinearity, it is often desirable to speed up or to slow
a marching scheme is used for the solution of the system, down the changes, from iteration to iteration, in the values
computationally they are much faster compared to elliptic of the dependent variable. This process is called over
systems; (c) Grid line spacing may be controlled by the cell laxation or under laxation depending on whether the
area or arc-length functions. variable changes are accelerated or slowed down: Over
The disadvantages are: (a) Boundary discontinuity may relaxation is often used in conjunction with the Gauss-
be propagated into the interior domain; (b) Specifying the Seidel method the resulting scheme being known as
cell-area or arc-length functions must be handled carefully. Successive Over-Relaxation (SOR). With the line-by-line
A bad selection of these functions easily leads to method, the use of over relaxation is less common. Under
undesirable grid systems. relaxation is a very useful device for nonlinear problems. It
is often employed to avoid divergence in the iterative
4.2. The Use of a Scheme solution of strongly nonlinear equations. There are many
ways of introducing over relaxation or under relaxation.
For certain specific values of the weighting factors fi, the
Some practices will be described here. We shall work with
discretization equation reduces to one of the well-known
the general discretization equation of the form
schemes for parabolic differential equations, [Patel and
Markatos, 1986]. In particular, f = 0 leads to the explicit
scheme, f = 0.5 to the Crank-Nicolson scheme, and f = 1 to (35)
the fully implicit scheme. Here, we present the equations of
the fully implicit scheme which is the most common in use. Another technique of over relaxation or under relaxation
is to replace the discretization equation (35) with the next
(29) equation where i is the so called inertia and a is the suitable
relaxation factor. For positive values of i the equation has
Where the effect of under relaxation and for negative values of i
produce over relaxation.
44 Anestis Stylianos: The Process of Heat Transfer and Fluid Flow in CFD Problems
(36)
4.4. Convection and Diffusion
So far we have seen how to formulate the discretization
equation from the general differential equation containing
the unsteady term, the diffusion term, and the source term.
The only omission has been the convection term, which we
shall now include. We have also dealt with the methods of
solving the algebraic equations; as long as the addition of
the convection term does not alter the form of the
discretization equation, the same methods continue to apply.
The convection is created by fluid flow. Our task is to
obtain a solution for in the presence of a given flow field
(i.e., the velocity components and the density). The origin
of the flow-field information is immaterial here. Having
somehow acquired the flow field, we wish to calculate the
temperature, concentration, enthalpy, or any such quantity
that is represented by the general variable . Although
convection is the only new term introduced in this chapter,
its formulation is not very straightforward. The convection
term has an inseparable connection with the diffusion term,
and therefore, the two terms need to be handled as one unit.
5. Specify Initial Values and
Boundary Conditions
The flow can be from the top left hand corner. The flow
field is computed by solving the mass and momentum
conservation equations in a boundary fitted coordinate
system along with a set of realistic boundary conditions.
The boundary conforms to a regular Cartesian system in
case of no inclined wall. The free surface of the liquid was
considered to be flat and the slag depth was considered to
be insignificant. With these two assumptions the flow field
was solved with the help of the above equations for all the
turbulence models. The effect of natural convection is
ignored because the ratio, Gr/Re2=0.044 [Lopez-
Ramirez et al., 2000], where T, the driving force for
natural convection is the temperature difference between
the liquid steel at the top free surface and the bulk
temperature of the liquid, which is much less than unity for
all the cases that are computed here.
The formation of waves at the free surface was ignored.
The free surface was assumed to be flat and mobile. Fluxes
of all quantities across the free surface were assumed to be
zero [Szekely et al,, 1987]. Therefore, normal velocity
component (for convective flux) and normal gradients of
all variables (for diffusive flux) were all set to zero, i.e.
(37)
The exit can be computationally treated as either a
standard outflow or as a plane or surface, at which flow
occurs at an ambient pressure (taken). At the outlets, both
types of boundary conditions were applied in order to
assess the similarity of the experimental results to model
configuration.
At all the solid walls, the velocity components was set to
zero, at both the side walls,
(38)
At both the frontal side walls,
(39)
At the bottom wall
(40)
6. Computer Program and CFD
Software
It should be noted that, while constructing the
discretization equations, we cast them into a linear form
but did not assume that a particular method would be used
for their solution. Therefore, any suitable solution method
can be employed at this stage. It is useful to consider the
derivation of the equations and their solution as two
distinct operations, and there is no need for the choices in
one to influence the other. In a computer program, the two
operations can be conveniently performed in separate
sections, and either section can be independently modified
when desired.
So far, we have obtained the multidimensional
discretization equations by a straightforward extension of
the one-dimensional situation. One procedure that cannot
so easily be extended to multiple dimensions is the
tridiagonal matrix algorithm (TDMA). Direct methods (i.e.,
those requiring no iteration) for solving the algebraic
equations arising in two- or three-dimensional problems
are much more complicated and require rather large
amounts of computer storage and time. For a linear
problem, which requires the solution of the algebraic
equations only once, a direct method may be acceptable;
but in nonlinear problems, since the equations have to be
solved repeatedly with updated coefficients, the use of a
direct method is usually not economical. We shall,
therefore, exclude direct methods from further
consideration, except to say that a computer program for
the direct solution of discretization equations in two
dimensions. There are many iterative methods for solving
algebraic equations.
6.1. The Tridiagonal Matrix Algorithm
(TDMA)
In numerical linear algebra, the tridiagonal matrix
algorithm, also known as the Thomas algorithm, is a
 American Journal of Science and Technology 2014, 1(1): 36-49
simplified
mplified form of Gaussian elimination that can be used to
solve tridiagonal systems of equations. A tridiagonal
system for n unknowns may be written as
Where and . algorithm, in which by changing the definition of the
For such systems, the solution can be obtained in
operations instead of required by Gaussian
elimination. A first sweep eliminates the 's, and then an
(abbreviated) backward substitution produces the solution.
Examples of such matrices commonly arise from the
discretization of 1D Poisson equation (e.g., the 1D
diffusion problems) and natural
tural cubic spline interpolation;
similar systems of matrices arise in tight binding physics or
nearest neighbor effects models.
6.2. The Algorithm for
or Numerical
Simulation
Inn the numerical solution of incompressible fluid flow
and heat transfer problems, the pressure-correction
pressure
approach is the most popular method used in CFD/NHT
community. The first pressure-correction
correction algorithm was the
SIMPLE proposed by Patankar and Spalding
Spaldin (1972). The
acronym SIMPLE stands for semi-implicit
implicit method for the
pressure-linked equation. The major approximations made
in the SIMPLE algorithm are (Tao, 2001):
The initial pressure field and the initial velocity fields
are independently assumed, hence henc the inherent
interconnection between pressure and velocities are
neglected, leading to some inconsistency between them.
The effects of the pressure corrections of the neighboring
grids are arbitrary dropped in order to simplify the solution
procedure, thus
us make the algorithm semi-implicit.
semi
These assumptions will not affect the final solutions if
the iterative process converges (Tao, 2001; Patankar, 1980).
However, they do affect the convergence rate. As described
in Shyy and Mittal (1998), the great simplicity
simpl of the
SIMPLE algorithm comes from the neglecting the terms
that couples neighboring velocity values in the equation for
the velocity correction. However, this can also cause slow
convergence of the SIMPLE algorithm and it has been
found this neglect tends to over predict the pressure
correction and under relaxation for the pressure correction
has to be resorted to in order to stabilize the iterative
procedure. Therefore, a number of variants SIMPLE
algorithms were proposed in order to overcome one or both
of the approximations (Liu et al., 2005).
).
The SIMPLER algorithm (Patankar, 1981) 1981 successfully
overcome the first approximation, and is widely used in
45
the current CFD/NHT community. Even though there are
more than ten variants of the SIMPLE-like
SIMPLE algorithm, the
second approximation, i.e. the drop of the neighboring grid
effects, have not been successfully resolved so far and
many attempts have been made to resolve the problem. Van
Doormaal and Raithby (1984) proposed the SIMPLEC
coefficients of the velocity correction equation the effects
of this drop is partially compensated. In the algorithm
SIMPLEX (van Doormaal and Raithby, 1985; Raithby and
Schneider, 1988), by solving a set of algebraic
algebrai equation for
the coefficients in the velocity correction equations, the
effects of dropping the neighboring grids are also taken
into account in some degree. In 1985, the PISO method is
proposed by Issa (1985) to implement two or more
correction steps of pressure correction. In 1986 Connell
and Stow (1986) proposed two variants of pressure
correction process. Chatwani and Turan (1991) proposed a
pressure-velocity
velocity coupling algorithm to determine the
under-relaxation
relaxation factor in the pressure correction equation
equati
based on the minimization of the global mass residual
norm. Lee and Tzong (1992) introduced an artificial source
term into the pressure-linked
linked equation to improve the
convergence performance. Yen and Liu (1993) proposed
the explicit correction step methodmeth to accelerate the
convergence by making the velocity explicitly satisfy the
momentum equation. For buoyancy driven fluid flows
Sheng et al. (1998) introduced a temperature correction
into the velocity correction equation. Yu et al.(2001)
modified the SIMPLER
MPLER algorithm by artificially changing
the under-relaxation
relaxation term to match the variable to be solved.
The revised method was called MSIMPLER. All the
above-mentioned
mentioned algorithms and some others not
mentioned above (for example, SIMPLESSEC,
SIMPLESSE of Gjesdal sdal and Lossius (1997), and the
method proposed in Wen and Ingham (1993) are usually
called SIMPLE-like like or SIMPLE-family
SIMPLE algorithm. The
character common to all these algorithms is that a pressure
correction term is introduced to the segregated solution
process
ocess to improve the velocity and the effects of the
pressure corrections of the neighboring grid points are
neglected. Moukalled and Darwish (2000) made a
comprehensive review and reorganization of the express
format for all the pressure correction algorithms.
algori It can be
seen that the SIMPLER algorithm successfully overcomes
the first approximation, while almost all other variants of
the SIMPLE algorithm concentrate on overcoming the
second approximation. There seems no such attempt in the
literature to combine
mbine the SIMPLER algorithm and one of
the other variants so that the effects of both of the two
approximations can be alleviated in a better degree in one
algorithm.
Liu et al., (2005) work, the idea of SIMPLEC was
incorporated into the SIMPLER algorithm to overcome the
second approximation in some extent. The revised
algorithm is called consistent-SIMPLER
consistent (CSIMPLER)
hereafter. Numerical experiments showed that CSIMPLER
46 Anestis Stylianos: The Process of Heat Transfer and Fluid Flow in CFD Problems
can generally accelerate the rate of convergence, especially
for natural convection and the cases with low under-
relaxation factor.
In the last few years there has been continuous progress
in the development of CFD codes. These codes can now
cope with a high level of complexity in many research
fields, which makes them attractive to use. However, CFD
codes that are used in the food processing industry have not
reached a relatively mature state, they still require to be
improved in accuracy, ease of use, robustness and
computational efficiency. Of all the commercial CFD codes
there are some codes for general purpose as well as for
specific applications (e.g. non-Newtonian flow). Most of
these commercial CFD codes can be supported on the
platforms of UNIX on workstations and WINDOWS or
LINUX on high-end Intel Pentium PCs. Table 2
summarizes the main commercial CFD codes.
Table 2. Commercial CFD code list.
CFD code Web site
CFX http://www.software.aeat.com/cfx/
FLUET http://www.fluent.com/
PHOENICS http://www.cham.co.uk/
STAR-CD http://www.cd.co.uk
FLOW-3D http://www.flow3d.com
CFD-ACE http://www.cfdrc.com
ICEM CFD http://www.icemcfd.com/icepak.html
TECPLOT http://www.amtec.com
PAM-FLOW http://www.esi.com.au/
FLOVENT http://www.flomerics.com/
6.2.1. The SIMPLER Algorithm
In all cases considered, the flow was assumed to be
Newtonian, laminar and two-dimensional. Viscous
dissipation is omitted. All thermo physical properties
except density are presumed constant. The Boussinesq
approximation is used for natural convection problem.
Discretizing the governing equations by the finite volume
method (Tao, 2001; Moukalled and Darwish, 2000) on a
staggered grid system (Figure 6), we have another form of
the equation (35)
Figure 6. Control volumes in 2D Cartesian coordinates.
Where is the general valuable standing for u and v, the
subscripts P and nb refer to the gird point P and its
neighboring grids, respectively, ap is the coefficient for the
main grid point, anbs are the coefficients of neighboring
grid points and b is the source term. For the discretized
momentum equations separating the pressure gradient term
from the b-term and replace the general valuable by u or v,
we have:
(42)
(43)
The discretized pressure equation is deduced from the
momentum equations and the continuity equation and can
be expressed as, (Tao, 2001; Patankar, 1980):
(44)
By solving equations (42) and (43), we can obtained
the intermediate solutions, symbolized by u * and v * which
need to be improved such that the improved velocities can
satisfy the mass conservation condition for each control
volume. By introducing a pressure correction term, p0 ,
and the corresponding velocity correction terms u and v ,
the improved velocities can be expressed by
(45)
(46)
These improved velocities are required to satisfy the
continuity condition. The equations for the velocity
correction terms, u, v, can be derived by some
substitution and rearrangement (Tao, 2001; Patankar,
1980), and take the following form:
(47)
(48)
At this point an approximation, i.e. the second
approximation in the SIMPLE algorithm mentioned above,
is introduced: dropping the terms and
in the above equations to simplify the
expressions. Then we obtain
(49)
where de, dn are defined as
(50)
Then the improved velocities are rewritten as follows:
 American Journal of Science and Technology 2014, 1(1): 36-49
(51)
Substitution of the improved velocities of equations
(50) and (51) into continuity equation, the equation for the
pressure correction term is then derived
(52)
where
(53)
In equation (53) the coefficients are the same as those in
the equation (16) except the b-term, where the velocities
take the values of the previous iteration, rather than the
intermediate solutions.
The solution procedure of the SIMPLER algorithm is as
follows:
guess a initial velocity field u 0, v 0;
calculate the coefficients of the discretized
momentum equations and the pseudo-velocities u^
and v^ by following equations:
(54)
solve pressure equation to get p *;
solve the discretized momentum equations with p * to
get u * and v *;
solve the pressure correction equation to get p0 ;
correct the velocities by equations (50) and (51);
solve the discretized equations for other scalar
variables if necessary; and
Return to step 2 until convergence condition is
satisfied.
It is to be noted that in the SIMPLER algorithm, the
pressure correction term is only used to correct the
velocities, but not used to correct the pressure. The
pressure correction values are over predicted by solving
equation (52), because the effects of the velocity corrections
at neighboring grid points are totally neglected. The
obtained pressure correction values are appropriate to
correct the velocities, but not to pressure values. Since the
discretized equations are all solved by iterative method,
the solutions of velocities of the current iteration are
based on the coefficients and source term determined by
the solutions of the last iteration. In particular, the
pressure field is solved according to the velocities of the
previous iteration, and it is in this aspect that the solved
velocity field and the pressure field are not consistent. It
is the authors consideration that the pressure filed may be
further revised within the iteration that the consistency
between the two fields can be refined.
47
6.2.2. The SIMPLEC Method
We obtain the following velocity correction equations:
(55)
where
(56)
The SIMPLEC algorithm alleviates in some degree the
effect of the second approximation in the SIMPLE
algorithm. The pressure correction equation in the
SIMPLEC algorithm is the same as that in the SIMPLE
algorithm except that the d-terms are calculated from
equation (50). The solution procedure of the SIMPLEC
algorithm is identical to that of SIMPLE (Tao,2001;
Patankar, 1980).
6.2.3. The CSIMPLER Algorithm
Now we incorporate the major idea of the SIMPLEC
algorithm into the SIMPLER algorithms as follows:
The pressure correction equation of the SIMPLEC is
adopted in the SIMPLER algorithm, i.e. the d-terms are
calculated from equation (56); the pressure is also
corrected after the pressure correction equation is solved:
(57)
Where ap is the relaxation factor for the pressure
correction. When ap<1 it is under relaxation of the pressure
correction, while ap>1 implies the over relaxation. Our
practices have shown that over relaxation of the pressure
correction term is often useful for the acceleration of the
convergence procedure which will be discussed later.
By adopting above two treatments into SIMPLER
algorithm and keeping the solution procedures the same,
the resulting solution algorithm is called consistent
SIMPLER, simplified by CSIMPLER, [Ashraya et al.,
2007].
We consider that the introduction of the pressure
correction term into the present pressure will improve the
coupling between velocity and pressure, hence, may
accelerate the convergence of the iterative process. It is to
be noted that for any existing code based on the SIMPLER
algorithm the implementation of the CSIMPLER algorithm
is very simple and easy.
7. Conclusion
The objective of this article is to illustrate all steps to
find the solution of a heat transfer and fluid flow problem
using CFD packages. In order to present them we gave
only the necessary information in theoretical fluid
dynamics. Searching, we can find may research papers and
48 Anestis Stylianos: The Process of Heat Transfer and Fluid Flow in CFD Problems

books specialize in this area. [7] Connell, S.D. and Stow, P. (1986), The pressure correction
methods, Comput. Fluids, Vol. 14, pp. 1-10.

Nomeclature [8] Donaldson C. du P., Calculation of turbulent shear flows for

atmospheric and vortex motions. AIAA J. 10, 4-12 (1972).
g: Acceleration due to gravity (m/s2) [9] Gjesdal, T. and Lossius, M.E.H. (1997), Comparison of
k: Thermal conductivity (W/m' K) pressure correction smoothers for multigrid solution of
p: Pressure (Pa) incompressible flow, Int. J. Numer. Methods Fluids, Vol.
Pr : Prandtl number 25, pp. 393-405.
Re: Reynolds number [10] Gosman A. D., W. M. Pun, A. K. Runchal, D. B. Spalding
Sc: Schmidt number and M. Wolfshtein, Heat and Mass Transfer in Recirculating
T: Temperature (K) Flows. Academic Press, London (1969).
t: time (s) [11] Hirsch Charles. 2007. Numerical computation of internal
u,v,w Velocity (m/s) and external flows. The fundamentals of Computational
: diffusion flux coefficient Fluid Dynamics. John wiley and Sons Publ. LTd.
R: rate of chemical generation [12] Hoffman Klaus A. and Steve J. Chiang. Computational
: density (kg/m3) Fluid Dynamics, volume II. By Engineering Education
: viscosity (kgm/ms) System, USA, 2000
: scalar variable
[13] Issa, R.I. (1985), Solution of implicitly discretized fluid
: dissipation rate of turbulent kinetic energy (m2/s3) flow equation by operator-splitting, J. Comput. Physics,
k: turbulent kinetic energy (m2/s2) Vol. 62, pp. 40-65.
c1,c2 constants in the turbulence model
[14] Lam G.K.G. and K. Bremhorst. 1981. A modified form of
c , , k empirical constants the k- model for predicting wall turbulence. ASME, J.
cp heat capacity (J/kgr k) Fluids Eng. Vol.103, pp. 456-460
r volume fraction
[15] Launder B. E. and D. B. Spalding, Turbulence models and
Prandtl number their applications to the prediction of internal flows. Heat
Subscripts and Superscripts Fluid Flow 2, 43-54 (1972).
eff: Sum of molecular and turbulent contributions
ref: Reference quantity [16] Launder B. E., Stress transport closures--into the third
generation. Syrup. Turbulent Shear Flows 3, 259-266 (1982).
i,j: the three spatial x,y,z direction
s south [17] Launder B. E., W. C. Reynolds, W. Rodi, J. Mathieu and D.
n north Jeandel, Turbulence Models and Their Applications, Vol. 2.
Eyrolles, Paris (1984).
e east
w west [18] Launder; B. E. and Spalding, D. B. (1974). The Numerical
p center point Computation of Turbulent Flow, Comp. Methods Appl.
Mech. Eng., val. 3, p. 269-289.
t top
b bottom [19] Lee, S.L. and Tzong, R.Y. (1992), Artificial pressure for
pressure-linked equation, Int. J. Heat Mass Transfer, Vol.
35, pp. 2705-16.
References
[20] Lekshminarayana B., Turbulence modeling for complex
[1] Acharya S., B.R. Baliga, K. Kari, J.Y. Murthy, C. Prakash flows. AIAA. paper AIAA-85 1652 (1985).
and S.P. Vanka. 2007. Pressure based finite volume methods
in Computational Fluid dynamics. J. Heat Trasfer, 129(4), [21] Liu X.L., W.Q. Tao and Y.L. He (2005). A simple method
pp. 407-424 for improving the SIMPLER algorithm for numerical
simulations of incompressible fluid flow and heat transfer
[2] Bradshaw P. and T. Cebeci, Engineering Calculation problems., Int. J. heat Mass Transfer, vol. 22, no. 8, pp.
Methods. Academic Press, London (1978). 921-939

[3] Bradshaw P., D. H. Ferris and N. P. Atwell, Calculation of
boundary-layer development using the turbulent energy
equation. J. Fluid Mech. 28, 593~i16 (1967).
[4] Bradshaw P., The understanding and prediction of turbulent
flow. Aeronaut. J. 76, 403-417 (1972).
[5] Cebeci T., K. C. Chang and P. Bradshaw, Solution of
hyperbolic system of turbulence-model equations by the
"Box" scheme. Comput. Meth. appl. Mech. Engng 22, 213-
227 (1980).
[22] Lopez-Ramirez S., R.D. Morales, J.A. Romero Serrano.
2000. Numerical simulation of the effects of buoyancy
forces and flow control devices on fluid flow and heat
transfer phenomena of liquid steel in a tundish. Num. Heat
Transf., A, vol.37, pp. 37-69
[23] Lumley J. L., Computational modeling of turbulent flows.
Adv. appl. Mech. 18, 123-176 (1978).
[24] Lumley J. L., Turbulence modeling. J. Appl. Mech. 50, pp.
1097-1103 (1983).

[6] Chatwani, A.U. and Turan, A. (1991), Improved pressure- [25] Marvin J. G, Turbulence modeling for computational
velocity coupling algorithm based on global residual norm, aerodynamics. AIAA J. 21, 941-955 (1983).
Numer. Heat Transfer. Part B, Vol. 20, pp. 115-23.
 American Journal of Science and Technology 2014, 1(1): 36-49
[26] Meller G. L. and T. Yamada, Development of a turbulence
closure model for geophysical fluid problems. Rev.
Geophys. space Phys. 20, 851-875 (1982).
[27] Mellor G. L, and H. J. Herring, A survey of the mean
turbulent field closure models. AIAA J. 11, 590-599 (1973).
[28] Monson, D.J., Seegmiller, H.L., McConnaughey and Chen,
Y.S. (1990), Comparison of Experiment with calculations
using curvature corrected zero and two equations
turbulence models for a two-dimensional U-duct, AIAA,
pp. 1390-1484.
[29] Moukalled, F. and Darwish, M. (2000), A unified
formulation of the segregated class of algorithm for fluid
flow at all speeds, Numer. Heat Transfer B, Vol. 37, pp.
103-39.
[30] Murphy J. D., Turbulence Modeling. NASA TM 85889
(1984).
[31] Murthy, J.Y., Minkowycz, W.J., Sparrow, E.M., and Mathur,
S.R., 2007. Survey of Numerical Methods, Handbook of
Numerical Heat Transfer, 2nd ed., eds., Minkowycz, W.J.,
Sparrow, E.M., and Murthy, J.Y., pp. 3-51, Wiley, New York.
[32] Patankar S.V. (1980) Numerical heat transfer and fluid
flow. By Hemisphere Publishing Corporation.
[33] Patankar, S. V. and Spalding, D. B. (1970). Heat and Mass
Transfer in Boundary Layers, 2d ed., Intertext, London.
[34] Patankar, S.V. (1981), A calculation procedure for two-
dimensional elliptic situations, Numer. Heat Transfer, Vol.
4, pp. 409-25.
[35] Patankar, S.V. and Spalding, D.B. (1972), A calculation
procedure for heat mass and momentum transfer in three
dimensional parabolic flows, Int. J. Heat Mass Transfer,
Vol. 15, p. 1787.
[36] Patankar, S.V. (1975). Numerical Prediction of Three
Dimensional Flows, in B. E. Launder (ed.), Studies in
Convection: Theory, Measurement and Applications, vol. 1,
Academic, New York.
[37] Patel M.K. and N.C. Markatos. An evaluation of eight
discretization schemes for two dimensional convection
diffusion equations. Int. J. Num. Meth. Fluids, vol. 6, pp.
129-154, (1986).
[38] Patel V. C., W. Rodi and G. Scheuerer, Turbulence models
for near wall and low Reynolds number flows: a review.
AIAA J. 23, 1308-1319 (1985).
[39] Raithby, G.D. and Schneider, G.E. (1988), Elliptic
system: finite difference method II, in Minkowycz, W.J.,
Sparrow, E.M., Pletcher, R.H. and Schneider, G.E. (Eds),
Handbook of Numerical Heat Transfer, Wiley, New York,
NY, pp. 241-89.
[40] Reynolds W. C., Computation of turbulent flows. Ann. Rev.
Fluid Mech. 8, 183-208 (1976).
49
[41] Rodi W., Examples of turbulence models for incompressible
flows. AIAA J. 20, 872-879 (1982).
[42] Rotta J. C., Prediction methods for turbulent flows. Von
Karman Institute for Fluid Dynamics, Lecture Series 76
(1975).
[43] Sheng, Y., Shoukri, M., Sheng, G. and Wood, P. (1998), A
modification to the SIMPLE method for buoyancy-driven
flows, Numer. Heat Transfer, Vol. 33, pp. 65-78.
[44] Shyy, W. and Mittal, R. (1998), Solution methods for
the incompressible Navier-Stokes equations, in Johnson,
R.W. (Ed.), The Handbook of Fluid Dynamics, CRC Press,
Boca Raton, FL, pp. 31.1-31.33.
[45] Spalding D. B., Turbulence Models, a Lecture Course.
Imperial College of Science and Technology (1982).
[46] Sunden B. 2013. On computational heat transfer in
industrial applications. ASME 2013 Heat Transfer Summer
Conf. Collocated with the ASME 2013 7th Int. Conf. on
Energy Sustainability and the ASME 2013 11th Int. Conf.
on Fuel Cell Science, Engineering and Technology, HT
2013
[47] Szekely J., O.J. Ilegbussi and N.El-Kaddah. 1987. The
mathematical modeling of complex fluid flow phenomena
in tundishes. Physicochemical Hydrodynamics, vol.9, pp.
453-472
[48] Tao, W.Q. (2001), Numerical Heat Transfer, 2nd ed., Xian
Jiaotong University, Xian, p. 211. van Doormaal, J.P. and
Raithby, G.D. (1984), Enhancement of SIMPLE method
for predicting incompressible fluid flows, Numer. Heat
Transfer, Vol. 7, pp. 147-63.
[49] van Doormaal, J.P. and Raithby, G.D. (1985), An
evaluation of the segregated approach for predicting
incompressible fluid flow, Paper 85-HT-9, ASME.
[50] Wen, X. and Ingham, D.B. (1993), A new method for
accelerating the rate of convergence of the SIMPLE-like
algorithm, Int. J. Numer Methods Fluids, Vol. 17, pp. 385-
400.
[51] Wilcox D. C. and M. W. Rubesin, Progress in turbulence
modeling for complex flow fields including the effect of
compressibility. NASA TP 1517 (1980).
[52] Yahkot Y. and Orszag S.A., 1992. Development of
turbulence models for shear flows by a double expansion
technique. Phys. Fluids A, vol. 4, pp. 1510-1520
[53] Yen, R.H. and Liu, C.H. (1993), Enhancement of the
SIMPLER algorithm by an additional explicit corrector
step, Numer. Heat Transfer B, Vol. 24, pp. 127-41.
[54] Yu, B., Ozoe, H. and Tao, W.Q. (2001), A modified
pressure-correction scheme for the SIMPLER method,
MSIMPLER, Numer. Heat Transfer B, Vol. 39, pp. 439-49.
Further reading.